Bug 119281 - boot failure with intel_iommu=on after 5f0a7f7614a9d99325ac8d618f1cdf7a3014287c
Summary: boot failure with intel_iommu=on after 5f0a7f7614a9d99325ac8d618f1cdf7a3014287c
Status: RESOLVED CODE_FIX
Alias: None
Product: Drivers
Classification: Unclassified
Component: PCI (show other bugs)
Hardware: Intel Linux
: P1 blocking
Assignee: drivers_pci@kernel-bugs.osdl.org
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-05-30 17:47 UTC by florian.ziesche+linux
Modified: 2016-10-26 21:44 UTC (History)
2 users (show)

See Also:
Kernel Version: 4.2-rc1+
Subsystem:
Regression: Yes
Bisected commit-id:


Attachments
4.6.2 dmesg with 5f0a7f7614a9d99325ac8d618f1cdf7a3014287c reverted (79.22 KB, text/plain)
2016-06-15 19:08 UTC, florian.ziesche+linux
Details
4.1.0/b63d80d1e01e949dbe469e1d72fc0b7e173dbdd8 dmesg (63.73 KB, text/plain)
2016-06-15 19:36 UTC, florian.ziesche+linux
Details
Possible fix (1.01 KB, patch)
2016-06-17 09:30 UTC, Joerg Roedel
Details | Diff

Description florian.ziesche+linux 2016-05-30 17:47:57 UTC
Can no longer boot with intel_iommu=on after this is applied.
Reverting this for 4.2 (and similar for 4.6) allows booting again, and I can confirm that iommu/VT-d is still functional after this.
-> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/iommu/intel-iommu.c?id=5f0a7f7614a9d99325ac8d618f1cdf7a3014287c

Hardware:
Gigabyte GA-X58A-UD5 (Ff1 BIOS)

Probably the same issue has been reported at: https://www.redhat.com/archives/vfio-users/2016-May/msg00183.html
This is with GA-EX58-UD5 and Asus Rampage II Gene hardware on kernel 4.2 and 4.4.
Comment 1 Joerg Roedel 2016-06-15 17:12:14 UTC
Can you please post dmesg right after boot with a kernel that has 5f0a7f7614a9d99325ac8d618f1cdf7a3014287c reverted?
Comment 2 florian.ziesche+linux 2016-06-15 19:08:26 UTC
Created attachment 220181 [details]
4.6.2 dmesg with 5f0a7f7614a9d99325ac8d618f1cdf7a3014287c reverted

Also building a kernel @ https://git.kernel.org/cgit/linux/kernel/git/joro/iommu.git/commit/?h=iommu-updates-v4.2&id=b63d80d1e01e949dbe469e1d72fc0b7e173dbdd8 right now (b/c I forgot to save it somewhere ...), which is the commit before 5f0a7f7614a9d99325ac8d618f1cdf7a3014287c. I'll add a dmesg of that later as well.
Comment 3 florian.ziesche+linux 2016-06-15 19:36:57 UTC
Created attachment 220191 [details]
4.1.0/b63d80d1e01e949dbe469e1d72fc0b7e173dbdd8 dmesg

Also removed Nvidia and AMD drivers this time, so less noise.
Comment 4 Joerg Roedel 2016-06-17 09:30:31 UTC
Created attachment 220421 [details]
Possible fix

Can you please test whether this patch fixes the problem?
Comment 5 florian.ziesche+linux 2016-06-17 11:11:20 UTC
Yep, it's working. Tested on 4.6.2.
Comment 6 Joerg Roedel 2016-06-20 11:04:28 UTC
(In reply to florian.ziesche+linux from comment #5)
> Yep, it's working. Tested on 4.6.2.

Thanks, patch is now upstream in Linus' tree and should land in the stable-trees for older kernels soon.
Comment 7 Bjorn Helgaas 2016-10-26 21:44:57 UTC
Fixed by a4c34ff1c029, which appeared in v4.7.

http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=a4c34ff1c029

Note You need to log in before you can comment on or make changes to this bug.